Detection and mapping of exotic weeds using UAS and machine learning: Bitou Bush Case Study

Why it matters

Bitou bush (Chrysanthemoides monilifera) is an introduced species that is outcompeting and smothering hosts of Australian coastal dune ecosystems. Since 1980, infestations in Northern NSW have been controlled, and outbreaks in QLD have been plummeted, with only isolated plants being found in the field today. Weed surveillance is currently performed on-ground twice a year, involving biosecurity personnel assessing densely vegetated areas in emu parades to maximise detection capabilities. However, the detection and mapping of weeds are tedious, time-consuming and expensive tasks. Highly dense vegetation restricts profoundly ground surveillance. Moreover, lack of visibility increases the risk of ignoring weeds in the surveys. As a result, alternative monitoring techniques are required to ensure eradication.

Aerial image of an infested area

Image segmentation (weeds in colour and background in grayscale)

Bitou Bush flowering
Bitou bush in coastal NSW

 

 

Overview

  • Design a system to detect bitou bush in coastal dunes using unmanned aircraft and machine learning classification algorithms towards the development of a flexible approach to monitoring and track similar weeds of interest in New South Wales and Queensland.
  • Evaluate approaches to assist with automatic detection of invasive species through eradication and management projects.
  • Build support from the community and local government for the use of automatic detection of invasive species for eradication and management projects.

Real world impact

  • Ground-based and airborne data collection campaigns on designated zones with bitou bush under standard and challenging topography at critical season periods during the year, in collaboration with biosecurity experts and UAV technicians.
  • Evaluation of cost-effectiveness of data acquistion processes that ensure optimal collection of representative data in a broad scale.
  • Design of a system capable of mapping surveyed areas showing locations and distribution of bitou bush and a data set of coordinates of bitou bush detections entered into corporate record management systems.
  • Delivery of progress reports on the review of predictive modelling approaches combined with the feasibility of using imaging sensors onboard unmanned aircraft for invasive weeds detection and mapping.

Publications

1.      Sandino, J.; Harris, S.; Trotter, P.; Shukla, A.; Gonzalez, F. Revolutionising the detection and mapping of exotic weeds using UASs and machine learning. In International Master Class in Plant Biosecurity, 14-26 January 2018, Denpasar, Indonesia. Link: https://eprints.qut.edu.au/116356/

Chief Investigators

Team

Other Team Members

  • Stacy Harris (Biosecurity Queensland)
  • Peter Trotter (Aspect UAV Imaging)
  • Al Sim (Aspect UAV Imaging)